Member-only story
How to Teach a Machine to Spot Microscopic Changes in Retinal Vessels
In a world where diabetes has become an increasingly common threat to public health, scientists and medical professionals are hunting for new ways to diagnose eye complications earlier and more precisely. Diabetic retinopathy (DR) stands out as one of the leading causes of preventable blindness globally. That fact alone highlights how timely and critical it is to develop efficient screening and treatment strategies. Early detection of minute blood vessel anomalies, called microaneurysms, is especially vital, because they are the first visible signs of non-proliferative diabetic retinopathy (NPDR). Now, a team of investigators has introduced an impressive breakthrough: they have trained cutting-edge deep learning networks, called “U-nets,” to spot these microaneurysms automatically in images from a non-invasive eye imaging technique known as optical coherence tomography angiography (OCTA). Their achievements, reported in Scientific Reports (Volume 14, Article 21520, 2024) by authors Lennart Husvogt, Antonio Yaghy, Alex Camacho, Kenneth Lam, Julia Schottenhamml, Stefan B. Ploner, James G. Fujimoto, Nadia K. Waheed, and Andreas Maier, could help usher in a new era of rapid, safe, and cost-effective DR screening.
